Adam
@adamcreates_
Followers
252
Following
66
Statuses
45
relentless builder and creative - previous work on generative ai, 3D personalization, programming education. style is the answer. @relignai
/
Joined September 2024
RT @relignai: deepseek r1 grpo bounty complete__ open source flywheel at work. evaluations complete, training runs in progress, benchmar…
0
6
0
@danielesesta curious to learn about the base models you're using and how they are using reasoning to answer complex questions
0
0
1
đź‘€
to relign__ to realign a base model into a reasoning model, effectively teaching it how to think in steps. see also: taking the red pill, a baby walking, pulling back the curtain, emerging from the cave, returning to zero, peering behind the veil. first framework benchmarks incoming__
1
0
4
open source flywheel: 1. work on difficult problems 2. difficult problems attract high-caliber talent 3. high-caliber talent drives development / attracts more high-caliber talent powered by $RELIGN bounty program
1
4
21
RT @relignai: new developer bounty__ create a complex medical reasoning task specification + verifier based on HuatuoGPT-o1. 250K $RELIG…
0
11
0
RT @relignai: the relign framework turns base models into reasoning models__ this allows all ai applications built on top of these models…
0
13
0
RT @relignai: roadmap__ phase i - launch__complete phase ii - develop a strong reasoning library | grpo, full docs, whitepaper, modulariz…
0
19
0
@dabit3 @aitruth_io @gmEthereum @newsconomist Hey @dabit3 , would love to present @relignai - an open source reinforcement learning framework. dropped you an email (:
1
0
2
RT @relignai: relign gitbook v1 live - covering concept, vision, mission, roadmap, use cases notable quote: relign is a step towards artif…
0
2
0
RT @relignai: git org updated - full project history online developer active on deepseek r1 bounty bounty program expanding first live codi…
0
14
0
advancing artificial intelligence and its applications is a direct consequence of how well models can reason. for specific tasks such as agent actions, base models cannot reason. so, in plain terms, RELIGN re-aligns a base model into a reasoning model. one that is capable of thinking in steps, much like a human would about a complex problem. the result is a more powerful model that improves performance of anything built on top of it - agents in particular. big underscore on the importance of reinforcement learning. a quote from the deepseek r1 paper: One of the most remarkable aspects of this self-evolution is the emergence of sophisticated behaviors as the test-time computation increases. Behaviors such as reflection—where the model revisits and reevaluates its previous steps—and the exploration of alternative approaches to problem-solving arise spontaneously. These behaviors are not explicitly programmed but instead emerge as a result of the model’s interaction with the reinforcement learning environment. This spontaneous development significantly enhances DeepSeek-R1-Zero’s reasoning capabilities, enabling it to tackle more challenging tasks with greater efficiency and accuracy. for a deeper read on reinforcement learning - ___
11
11
57
first bounty out. go alone, go fast. go together, go far. ___
new developer bounty implement deepseek r1's GRPO in RELIGN and evaluate it on gsm8k math. 250K $RELIGN
1
2
10